kpcasvm.m

粒子群算法

function [maxerror,meanerror,stderror] = KPCASVM(T,G,C,P)


%UNTITLED1 Summary of this function goes here
%  Detailed explanation goes here

% T= 77.7149       
% G=18.6755
% C=65.8118 
% P=64.6552

b=f2b(T,[0 10],8);
% tic;
% H1=xlsread('tading.xls',1,'A1:H123');
H=xlsread('testy.xls',1,'A1:J60');
H1=[];
for L=1:8
if b(L)==1
    H1=[H1,H(:,L)];

end
end
H1=[H1,H(:,9:10)];  %加上最好两列输出值

len=size(H1,1);
Vol=size(H1,2);
H2=H1(:,Vol-1);

delta=(max(H2)-min(H2));  %反归一
delta1=min(H2); 


%  [b,a]= butter(5,0.8);
%  H1=filtfilt(b,a,H);
%   
% 
% 
%  k=0;
% P=zeros(len,9);
%  for j=1:9
%     delta= std(H1(:,j));
%     avg=mean(H1(:,j));
%      for i=1:len-k
%          if i>len-k
%              break
%          end
%          temp= H1(i,j)-avg;
%          if temp>=3*delta
%              P(i,j)=1;
%              H1=[H1(1:i-1,:);H1(i+1:len-k,:)];
%             
%              k=k+1;
%          end
%      end
%  end
%  [i,j]=find(P)
 
 data=H1;
 data=(data-ones(size(data,1),1)*min(data))./(ones(size(data,1),1)*(max(data)-min(data)));
  H1=data-ones(size(data,1),1)*mean(data);

 
 k=1;
 for L=1:len
     
    
     switch L
         case 1
             A1=H1(2:len,1:Vol-2);
             B1=H1(2:len,Vol-1);
             C1=H1(1,1:Vol-2);
             D1=H1(1,Vol-1);
         case len
             A1=H1(1:len-1,1:Vol-2);
             B1=H1(1:len-1,Vol-1);
             C1=H1(len,1:Vol-2);
             D1=H1(len,Vol-1);

         otherwise
             A1=[H1(1:L-1,1:Vol-2);H1(L+1:len,1:Vol-2)];
             B1=[H1(1:L-1,Vol-1);H1(L+1:len,Vol-1)];
             C1=H1(L,1:Vol-2);
             D1=H1(L,Vol-1);



     end
 
 
 
%  A1=H1(1:40,1:8);
%  B1=H1(1:40,9);
%  C1=H1(41:60,1:8);
%  D1=H1(41:60,9);
 
      
     % max_ev =14;  %选择主元个数 tading
       max_ev =21;  %选择主元个数
      rbf_var = 1.3;  %髙斯函数的参数
      cov_size=size(A1,1);     %样本个数
      for i=1:cov_size,    %得到K
          for j=i:cov_size,
              K(i,j) = exp(-norm(A1(i,:)-A1(j,:))^2/rbf_var);
              K(j,i) = K(i,j);
          end
      end
      unit = ones(cov_size, cov_size)/cov_size;
      % centering in feature space!
      K_n = K - unit*K - K*unit + unit*K*unit;

      [evecs,evals] = eig(K_n);
     % A2=evecs;
      evals = real(diag(evals));
      for i=1:cov_size,
          evecs(:,i) = evecs(:,i)/(sqrt(evals(i)));   
      end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
      test_num=size(A1,1);

      unit_test = ones(test_num,cov_size)/cov_size;
      K_test = zeros(test_num,cov_size);
      for i=1:test_num,
          for j=1:cov_size,
              K_test(i,j) = exp(-norm(A1(i,:)-A1(j,:))^2/rbf_var);
          end
      end
      K_test_n = K_test - unit_test*K - K_test*unit + unit_test*K*unit;
      test_features = zeros(test_num, max_ev);
      test_features = K_test_n * evecs(:,1:max_ev);
      A2=test_features;


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
      test_num=size(C1,1);
      unit_test = ones(test_num,cov_size)/cov_size;
      K_test = zeros(test_num,cov_size);
      for i=1:test_num,
          for j=1:cov_size,
              K_test(i,j) = exp(-norm(C1(i,:)-A1(j,:))^2/rbf_var);
          end
      end
      K_test_n = K_test - unit_test*K - K_test*unit + unit_test*K*unit;
      test_features = zeros(test_num, max_ev);
      test_features = K_test_n * evecs(:,1:max_ev);
      C2=test_features;

      
%       model = svmtrain(B1, A2, '-s 3 -t 2 -g 0.11 -c 50 -p 0.05');
%   model = svmtrain(B1, A2, '-s 3 -t 2 -g 0.9876  -c 21.1079 -p 1.1029 ');
% G=0.11;
% C=50;
% P=0.05;
      str= ['-s 3 -t 2 -g ', num2str(G),  ' -c ' num2str(C) ' -p ' num2str(P)];
      model = svmtrain(B1, A2, str);
      [y, accuracy] = svmpredict(D1, C2, model);
      
%        y= y*delta+delta1;   
%        if L==9
%        y=y-8;
%        end
%        x=abs(H2(L)-y)/H2(L);
%        x=abs((D1-y)/D1);
      if k==1
          Y=y;
%           X=x;
          D=D1;
          k=0;
      else
          Y=[Y;y];
%           X=[X;x];
          D=[D;D1];
      end
      

 end
%   if  size(Y,1)==0
%            G
%            C
%            P
%        end
 Y=Y+mean(data(:,Vol-1));
%  D=data(:,9);
 Y= Y*delta+delta1;
%  D= D*delta+delta1;
 X=abs(H2-Y)./H2;

%  X=abs(D-Y)./D;
% toc;
% t=toc;
meanerror=mean(X);
maxerror=max(X);
stderror=std(X);


% save('error.mat','meanerror','maxerror','stderror');
%  X=abs(D1-y);
%  error1=sqrt(sum(X.^2)/test_num)
%  error2=max(abs(D1-y))
 

 
% Y=ones(size(X,1),1)*mean(X);
 %error=sqrt(sum((X-Y).^2)/test_num)
%  
%  y= y*(max(H2)-min(H2))+min(H2);   %反归一
 
%  X=abs(D2-y);
%  error1=sqrt(sum(X.^2)/test_num)
%  error2=max(abs(D2-y))
%  
%  T1=[50:1:100]';
%  T2=[50:1:100]';
% %  V=[1,60,0,0.2];
% plot(H2,Y,'bd','LineWidth',2 );
% hold on
% plot(T1,T2, 'k-','LineWidth',2 );
% hold on
% ylabel('实际值\bf%');
% xlabel('测量值\bf%');
% plot(T1+5,T2, 'k-','LineWidth',2 );
% axis(V);
%  plot(H2,'r--','LineWidth',2 );
% %  axis(V);
%  hold on
%  plot(Y,'b','linewidth',2);
%  axis(V);